Beam team says model was trained from scratch, not distilled from closed models
alexpolozov · x · 2026-10-07
Responding to community speculation that Beam might be distilling from closed models, team member Chetan Tekur clarified that Beam was trained entirely from scratch, with no distillation from closed models, pointing to an earlier explanation by Misha. The retweet by alexpolozov (ex-DeepMind) adds weight to the clarification amid ongoing debate over the provenance of newly released models.
More from Companies & People
- Insider Predicts Meta Could Soon Turn Into a Top Frontier-Model Player — BLUECOW009 · 2026-10-07
- Sam Altman: 'We are entering a new era of discovery now' — sama · 2026-10-07
- Niessner's SpAItial opens 15 roles across world models, 3D diffusion and robot learning — CSProfKGD · 2026-10-07
- 1X NEO Hand Engineer Max Titov Goes Independent, Shares Robotics Portfolio — eigenron · 2026-10-07
- OpenAI Drops 700 Preprints of Mathematical Proofs and Counterexamples — tootie · 2026-10-07
- Heaviest AI release day in recent history: OpenAI math manuscripts, Agent Protocol, Le Chonk — altryne · 2026-10-07